Stefano MarinelliMay 10

Happy Sunday, #BSDCafe
Happy Sunday, #illumosCafe
Happy Sunday, #Fediverse

I'm dismantling a small Kubernetes cluster and converting everything into FreeBSD jails. From 3 VMs to 1 smaller one, better performance, easier maintenance.

Do we really need a truck to carry a single pizza?

#RunBSD #FreeBSD

Show threadTomMay 10
@stefano Would love to read a write up on it without giving sensitive information out. I love reading about these types of stories. Do you also tailor the recordsize on the zfs datasets as well?
Show threadStefano MarinelliMay 10

@pertho I've written something like this in the past, but I'm taking notes for a future ITNotes blog post 😉

For ZFS - it depends. I usually don't do it, but in some specific use cases I've done it. I can't say if it was useful or not.

Show threadTom
@stefano Something like 16k for mysql data is supposed to be good, plus turning off mysql double writes.
May 10 at 9:39pmWeb
Show threadStefano MarinelliMay 11
@pertho it usually does and yes, double writes are bad there.
Show threadIan WagnerMay 11
@stefano @pertho no experience with MySQL in the past 15 years, but it definitely can make a difference for heavy Postgres workloads. And of course a large record size is optimal if the dataset is a bunch of large files.